Download Improving Customer Relationship Management Using Data Mining

Survey
yes no Was this document useful for you?
   Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Document related concepts

Green marketing wikipedia, lookup

Revenue management wikipedia, lookup

Bayesian inference in marketing wikipedia, lookup

Touchpoint wikipedia, lookup

Integrated marketing communications wikipedia, lookup

Digital marketing wikipedia, lookup

Marketing mix modeling wikipedia, lookup

Marketing wikipedia, lookup

Advertising campaign wikipedia, lookup

Visual merchandising wikipedia, lookup

Street marketing wikipedia, lookup

Segmenting-targeting-positioning wikipedia, lookup

Target market wikipedia, lookup

Global marketing wikipedia, lookup

Marketing strategy wikipedia, lookup

Product planning wikipedia, lookup

Direct marketing wikipedia, lookup

Retail wikipedia, lookup

Customer experience wikipedia, lookup

Sensory branding wikipedia, lookup

Services marketing wikipedia, lookup

Customer satisfaction wikipedia, lookup

Customer engagement wikipedia, lookup

Service blueprint wikipedia, lookup

Customer relationship management wikipedia, lookup

Transcript
International Journal of Machine Learning and Computing, Vol. 2, No. 6, December 2012
Improving Customer Relationship Management
Using Data Mining
Gaurav Gupta and Himanshu Aggarwal
Abstract—Customer Relationship Management (CRM)
refers to the methodologies and tools that help businesses
manage customer relationships in an organized way. Data
Mining is the process that uses a variety of data analysis and
modeling techniques to discover patterns and relationships in
data that may be used to make accurate predictions. It can
help you to select the right prospects on whom to focus. The
essence of the information technology revolution and, in
particular, the World Wide Web is the opportunity to build
better relationships with customers than has been previously
possible in the offline world. For instance, as part of their
CRM strategy, a business might use a database of customer
information to help construct a customer satisfaction survey,
or decide which new product their customers might be
interested in.
Index Terms—CRM, LCV, Marketing, Relationship, Data
Mining.
I. INTRODUCTION
Customer Relationship Management (CRM)[1] refers to
the methodologies and tools that help businesses manage
customer relationships in an organized way. CRM simply
means managing all customer interactions which requires
using information[15] about your customers and prospects
to more effectively interact with your customers in all
stages of your relationship with them.
The essence of the information technology revolution and,
in particular, the World Wide Web is the opportunity to
build better relationships with customers than has been
previously possible in the offline world. Until recently most
CRM software has focused on simplifying the organization
and management of customer information. Such software,
called operational CRM, has focused on creating a customer
database that prevents a consistent picture of customer’s
relationship with the company, and providing the
information in specific applications. However, the sheer
volumes of customer information and increasingly complex
interactions with customers have propelled data mining to
the forefront of making your customer relationship
profitable.
Data Mining is the process that uses a variety of data
analysis and modeling techniques to discover patterns and
relationships in data that may be used to make accurate
predictions. It can help you to select the right prospects on
whom to focus, offers the right additional products to your
Manuscript received July 9, 2012. Revised September 21, 2012.
F. Gaurav Gupta, Assistant Professor,is with the RIMT-Institute of
Engineering & Technology, Punjab, India (e-mail: [email protected]
yahoo.com).
S. Himanshu Aggarwal, Associate Professor, is with the University
College of Engineering, Punjabi University, Patiala, India. (e-mail:
[email protected]).
10.7763/IJMLC.2012.V2.256
existing customers and identify good customers who may be
about to leave. CRM applications that use data mining are
called analytic CRM.
By combining the abilities to respond directly to
customer requests and to provide the customer with a highly
interactive, customized experience, companies have a
greater ability today to establish, nurture, and sustain longterm customer relationships than ever before. The ultimate
goal is to transform these relationships into greater
profitability by increasing repeat purchase rates and
reducing customer acquisition costs. The needs to better
understand customer behavior and focus on those customers
who can deliver long-term profits has changed how
marketers [3] view the world.
Traditionally, marketers have been trained to acquire
customers, either new ones who have not bought the
product category before or those who are currently
competitors’ customers. This has required heavy doses of
mass advertising and price-oriented promotions to
customers and channel members. Today, the tone of the
conversation has changed from customer acquisition to
retention. A good thought experiment for an executive
audience is to ask them how much they spend and/or focus
on acquisition versus retention activities. While it is
difficult to perfectly distinguish between the two activities
although, the answer is usually that acquisition dominates
retention.
II. CRM MODEL
A basic model of CRM contains a set of 7 basic
components as shown in Fig. 1.
Fig. 1. CRM model
874
International Journal of Machine Learning and Computing, Vol. 2, No. 6, December 2012
A. Creating a Customer Database
A necessary first step to a complete CRM solution is the
construction of a customer database or an information
file[4]. This serves as the basis for any CRM activity. For
Web-based businesses, this should be a relatively
straightforward task as the customer transactions and
contact information are accumulated as a natural part of the
interaction with customers. For existing companies that
have not previously collected much customer information,
the task will involve seeking historical customer contact
data from internal sources such as accounting and customer
service.
The nest pertinent question is: What should be collected
for the database? Ideally, the database should contain
information about the following:
• Transactions: This should include a complete purchase
history
• Customer contacts: Today, there are an increasing number
of customer contact points from multiple channels and
contexts.
• Descriptive information: This is for segmentation and
other data analysis purposes.
• Response to marketing stimuli: This part of the
information file should contain whether or not the customer
responded to a direct marketing initiative, a sales contact, or
any other direct contact.
segments provide information for deploying the
marketing[12] tools.
If individual customer-based profitability is also available
through LCV or similar analysis, it would seem to be a
simple task to determine on which customers to focus. The
marketing manager can use a number of criteria such as
simply choosing those customers that are profitable. The
goal is to use the customer profitability analysis to separate
customers that will provide the most long-term profits from
those that are currently hurting profits. This allows the
manager to “fire” customers that are too costly to serve
relative to the revenues being produced.
On what basis should these customer selection decisions
be made? One approach would be to take the current
profitability. An obvious problem is that by not accounting
for a customer’s possible growth in purchasing, this could
lead to applied eliminating a potentially important customer.
Therefore, the customers with high LCV could be chosen;
This will allow better job incorporating potential purchases.
However, these are difficult to predict and you could
include a large number of unprofitable customers in the
selected group. No matter what criterion is employed, deselected customers need to be chosen with care.
D. Targeting the Customers
Mass marketing approaches such as television, radio, or
print advertising are useful for generating awareness and
achieving other communications objectives, but they are
poorly-suited for CRM due to their impersonal nature. More
conventional approaches for targeting selected customers
include a portfolio of direct marketing methods such as
telemarketing, direct mail, and, when the nature of the
product is suitable, direct sales.
In particular, the new mantra, “1-to-1” [10] marketing [6],
has come to mean using the Internet to facilitate individual
relationship building with customers. An extremely popular
form of Internet-based direct marketing is the use of
personalized e-mails[13].
B. Analyzing the Data
Traditionally, customer databases have been analyzed[8]
with the intent to define customer segments. A variety of
multivariate statistical methods ranging such as cluster and
discriminate analysis have been used to group together
customers with similar behavioral patterns and descriptive
data which are then used to develop different product
offerings or direct marketing campaigns[5]. Direct marketers
have used such techniques for many years. Their goals are
to target the most profitable prospects for catalogue
mailings and to tailor the catalogues suitable for different
customers.
Lifetime Customer Value (LCV)[9] is a term that gives
an idea that each row/customer of the database should be
analyzed in terms of current and future profitability to the
firm. When a profit figure can be assigned to each customer,
the marketing manager can then decide which customers to
target. Other kinds of data analyses besides LCV are
appropriate for CRM purposes. Marketers are interested in
what products are often purchased together, often referred
to as market basket analysis.
E. Relationship Programs
While customer contact through direct e-mail offerings is
a useful component of CRM. It is more of a technique for
implementing CRM than a program itself. Relationships are
not built and sustained with direct e-mails themselves but
rather through the types of programs that are available for
which e-mail may be a delivery mechanism.
The overall goal of relationship programs[11] is to
deliver a higher level of customer satisfaction. Managers
today feel that customers match realizations and
expectations of product performance, and that it is critical
for them to deliver such performance at higher and higher
levels as expectations increase due to intense or heightened
competition, and changing customer needs. In addition,
research has shown that there is a strong, positive
relationship between customer satisfaction and profits[14].
Managers must constantly measure satisfaction levels and
develop programs that help to deliver performance beyond
targeted customer expectations. Fig. 2 shows the customer
retention program which allows the managers to retains and
makes good relation with customers.
C. Customer Selection[7]
After the database has been created and analyzed, the
next step is to consider which customers to target? The
results from the analysis could be of various types. If
segmentation-type analyses are performed on purchasing or
related behavior, the customers in the most desired
segments (e.g., highest purchasing rates, greatest brand
loyalty) would normally be selected first. Other segments
could also be chosen depending upon additional factors. For
example, if the customers in the heaviest purchasing
segment already purchase at a rate that implies further
purchasing is unlikely, a second tier with more potential
would also be attractive. The descriptor variables for these
875
International Journal of Machine Learning and Computing, Vol. 2, No. 6, December 2012
exchanging product-related information and to create
relationships between the customers and the company or
brand. These networks and relationships are called
communities. The goal is to take a prospective relationship
with a product and turn it into something more personal. In
this way, the manager can build an environment that makes
it more difficult for the customer to leave the “family” of
other people who also purchase from the company.
6) Privacy issues
The CRM system described here depends upon a
database of customer information and analysis of the data
for more effective targeting of marketing communications
and relationship-building activities. There is an obvious
tradeoff between the ability of companies to better deliver
customized products and services and the amount of
information necessary to enable this delivery. Particularly
with the popularity of the Internet, many consumers and
advocacy groups are concerned about the amount of
personal information that is contained in databases and how
it is being used.
Fig. 2. Customer retention program
1) Customer Service
Because customers have more choices today and the
targeted customers are most valuable to the company,
customer service must receive a high priority within the
Company. Programs designed to enhance customer service
are normally of two types. Reactive service is where the
customer has a problem (product failure, question about a
bill, product return) and contacts the company to solve it.
Most companies today have established infrastructures to
deal with reactive service situations. Proactive service is a
different matter; this is a situation where the manager has
decided not to wait for customers to contact the firm but to
rather be aggressive in establishing a dialogue with
customers prior to complaining or other behavior sparking a
reactive solution.
7) Metrics
The increased attention paid to CRM means that the
traditional metrics used by managers to measure the success
of their products and services in the marketplace have to be
updated. Financial and market-based indicators like
profitability, market share, and profit margins have been
and will continue to be important. However, in a CRM
world, increased emphasis is being placed on developing
measures that are customer-centric and give the manager a
better idea of how her CRM policies and programs are
working.
2) Loyalty/Frequency programs
Loyalty[2] programs (also called frequency programs)
provide rewards to customers for repeat purchasing. A
number of Web-based companies providing incentives for
repeat visits to Web sites. Although these have not been
wildly successful, it is clear that the price orientation of
many Web shoppers creates the need for programs that can
generate loyal behavior.
III. APPLYING DATA MINING TO CRM
In order to build good model of CRM system, there are a
number of steps that need to be followed. Following are the
basic steps of data mining for effective CRM: y
Define Business Problem
y
Build Marketing Database
y
Explore Data
y
Prepare Data For Modeling
y
Build Model
y
Evaluate Model
y
Deploy Model and Results
3) Customization
The notion of mass customization goes beyond 1-to-1
marketing as it implies the creation of products and services
for individual customers, not simply communicating to
them. Customization is termed “versioning.” It is, of course,
easier to do this for services and intangible information
goods than for products but the examples above show that
even manufacturers can take advantage of the increased
information available from customers to tailor products that
at least give the appearance of being customized even if
they are simply variations on a common base.
A. Define the Business Problem
Each CRM application will have one or more business
objectives for which there are a need to build an appropriate
model. Depending upon the goals, build a very different
model.
4) Reward programs
Different companies used to give rewards to its
customers on successive purchasing of their goods. Reward
can be a gift item or cash back or like discount coupon,
which can be applied on purchasing the next item from the
same company.
B. Build a Marketing Database
Next step is to build a marketing database because of
operational databases and corporate data warehouse. This
will often not contain the data in the form it is needed. Also
CRM applications may interfere with the speedy and
effective execution of these systems. After this there is a
need to clean and integrate the data.
5) Community Building
One of the major uses of the Web for both online and
offline businesses is to build a network of customers for
C. Explore the Data
Before building a good predictive model, it is important
876
International Journal of Machine Learning and Computing, Vol. 2, No. 6, December 2012
to understand your data. Graphing and visualization tools
are a vital aid in data preparation and their importance to
effective data analysis cannot be overemphasized.
see improvements in how companies work to establish longterm relationships with their customers. However, there is a
big difference between spending money on these people and
products and making it all work: implementation of CRM
practices is still far short of ideal. Everyone has his or her
own stories about poor customer service and emails sent to
companies without hearing a response.
We can expect that the technologies and methodologies
employed to implement the steps will improve as they
usually do. More companies are recognizing the importance
of creating databases and getting creative at capturing
customer information. Real-time analyses of customer
behavior on the Web for better customer selection and
targeting is already here which permits companies to
anticipate what customers are likely to buy. Companies will
learn how to develop better communities around their
brands giving customers more incentives to identify
themselves with those brands and exhibit higher levels of
loyalty.
D. Prepare Data for Modeling
This is the final data preparation step before building
models. There are four main parts of this step:
• Select the variables on which to build the model.
• Construct new predictors derive from the raw data
• Decide to select a subset or sample of your data on
which to build models
• Transform variables in accordance with the
requirements of the algorithm you choose for
building your model.
E. Data Mining Model Building
The most important phase in this is an iterative process.
There is a need to explore alternative models to find the one
that is most useful in solving your business problem. Most
CRM applications are based on a protocol called supervised
learning.
REFERENCES
[1]
F. Evaluate Your Results
Perhaps the most overrated metric for evaluating your
results is accuracy. Another measure that is frequently used
is lift.
[2]
[3]
G. Deploy Model and Results
In building a CRM application, data mining is often a
small but critical part of the final product. The way data
mining is actually built into the application is determined by
the nature of the customer interaction. There are two ways
you interact with your customers: they contact you (inbound)
or you contact them (outbound). The deployment
requirements are quite different.
[4]
[5]
[6]
[7]
[8]
IV. CONCLUSION
CRM is essential to compete effectively in today’s world.
The more effectively you can use the information about
your customers to meet their needs the more profitable you
will be. But operational CRM needs analytical CRM with
predictive data mining models. The route to a successful
business requires a marketing manager that understands its
customers and their requirements and implements data
mining with a good different model.
[9]
[10]
[11]
[12]
[13]
V. THE FUTURE OF CRM
[14]
With the increased penetration of CRM philosophies in
organizations and the concomitant rise in spending on
people and products to implement them, it is clear we will
877
Frederick F. Reichheld, The Loyalty Effect, December 2007
(Cambridge, MA: Harvard Business School Press).
Rashi Glazer, “Winning in Smart Markets,” Sloan Management
Review, summer, 1999.
Rashi Glazer, “Measuring the Knower: Towards a Theory of
Knowledge Equity," California Management Review vol. 40, no. 3,
Spring 1998, pp.175-194.
Michel Wedel and Wagner A. Kamakura, Market Segmentation:
Conceptual and Methodological Foundations, 2nd edition, New York:
Kluwer Academic Publishers, 1999.
Don Peppers and Martha Rogers, The One to One Future, (New York:
Doubleday, 1993), Ch.4.
Valarie A. Zeithaml, Roland T. Rust, and Katherine N. Lemon
elaborate on this section in their paper, “The Customer Pyramid:
Creating and Serving Profitable Customers,” vol. 43, no. 4 summer
2001.
Wendy W. Moe and Peter S. Fader describe click stream analysis
more completely in their paper, Uncovering Patterns in
Cybershopping,” vol. 43, no. 4, summer 2001.
For other formulas, see Paul D. Berger and Nada I. Nasr, “Customer
Lifetime Value: Marketing Models and Applications,” Journal of
Interactive Marketing, winter, 1998, pp.17-30.
Susan Kuchinskas, “One-to-one?” Business 2.0, September 12, 2000,
pp.141-8
Richard L. Oliver, Satisfaction: A Behavioral Perspective on the
Consumer, (Boston, MA: Irwin McGraw-Hill, 1997), and Valarie A.
Zeithaml and Mary Jo Bitner, Services Marketing, (Boston, MA:
Irwin McGraw-Hill, 2000) are good examples.
Seth Godin, Permission Marketing, (New York: Simon & Schuster,
1999).
Jim Nail, “The Email Marketing Dialogue,” Forrester Research Inc.,
January 2000.
James Cigliano, Margaret Georgiadis, Darren Pleasance, and Susan
Whalley, “The Price of Loyalty,” The McKinsey Quarterly, number 4,
2000.
Carl Shapiro and Hal R. Varian, Information Rules, Cambridge, MA:
Harvard Business School Press, 1999.